Reducing false positives in computer-aided detection
View Patent ↗Methods, systems, and related computer program products for computer-aided detection (CAD) of anatomical abnormalities in digital (or digitized) x-ray mammograms are described. The inventive techniques are based on using a foundational CAD processing algorithm that is characterized by at least one of non-shift-invariance, non-rotational-invariance, and non-inversional-invariance. According to one preferred embodiment, a first x-ray mammogram image of a breast is received, and at least one altered version thereof is generated that differs therefrom by at least one of image shift, image rotation, and image inversion. The first x-ray mammogram image and each of the at least one altered versions thereof are individually processed using the foundational CAD algorithm to generate a respective plurality of individual CAD detection sets. The plurality of CAD detection sets are then compared to generate an overall CAD detection set.
1. A method for computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a foundational CAD processing algorithm, said foundational CAD processing algorithm being characterized by at least one of non-shift-invariance, non-rotational-invariance, and non-inversional-invariance, the method comprising:
receiving a first x-ray mammogram image of a breast in digital form;
processing by digital computer said first x-ray mammogram image to generate at least one altered version thereof that differs therefrom by at least one of image shift, image rotation, and image inversion, wherein at least one of said altered versions comprises a spatially inverted version of said first x-ray mammogram image;
individually processing said first x-ray mammogram image and each of said at least one altered versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
2. The method of claim 1 , wherein said comparing comprises, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, including that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
3. The method of claim 2 , wherein said collective evaluation of outcomes for said breast location comprises a voting technique in which said breast location is included in said overall CAD detection set only if at least seventy percent (70%) of said CAD detection sets have identified said breast location for CAD marking, and wherein the method further comprises displaying said overall CAD detection set to a viewer in conjunction with said first x-ray mammogram image on an output display.
4. The method of claim 1 , wherein said at least one altered version includes at least ten (10) successively rotated versions of said first x-ray mammogram image having rotation angles distributed between about −5 degrees and 5 degrees.
5. The method of claim 1 , said first x-ray mammogram image having a number of pixels in a range of about 1K×1K to about 10K×10K pixels, wherein said at least one altered version includes at least ten (10) successively shifted versions of said first x-ray mammogram image having shift amounts distributed between about 2 pixels and 20 pixels.
6. A method for computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a foundational CAD processing algorithm, said foundational CAD processing algorithm being characterized by at least one of non-shift-invariance, non-rotational-invariance, and non-inversional-invariance, the method comprising:
receiving a first x-ray mammogram image of a breast in digital form;
processing by digital computer said first x-ray mammogram image to generate at least one altered version thereof that differs therefrom by at least one of image shift, image rotation, and image inversion, wherein at least one of said altered versions comprises a spatially inverted version of said first x-ray mammogram image, and (i) said first x-ray mammogram image comprises a CC view of the breast and said at least one altered version is inverted vertically in comparison to said first x-ray mammogram image;
individually processing said first x-ray mammogram image and each of said at least one altered versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
7. The method of claim 6 , wherein said comparing comprises, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, including that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
8. The method of claim 6 , wherein said collective evaluation of outcomes for said breast location comprises a voting technique in which said breast location is included in said overall CAD detection set only if at least seventy percent (70%) of said CAD detection sets have identified said breast location for CAD marking, and wherein the method further comprises displaying said overall CAD detection set to a viewer in conjunction with said first x-ray mammogram image on an output display.
9. The method of claim 6 , wherein said at least one altered version includes at least ten (10) successively rotated versions of said first x-ray mammogram image having rotation angles distributed between about −5 degrees and 5 degrees.
10. The method of claim 6 , said first x-ray mammogram image having a number of pixels in a range of about 1K×1K to about 10K×10K pixels, wherein said at least one altered version includes at least ten (10) successively shifted versions of said first x-ray mammogram image having shift amounts distributed between about 2 pixels and 20 pixels.
11. A non-transitory computer readable medium encoded with a computer program which when implemented by at least one processor causes the computer to implement computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a non-inversionally-invariant foundational CAD processing algorithm which performs the steps of:
receiving a first x-ray mammogram image of a breast in digital form;
processing said first x-ray mammogram image to generate at least one spatially inverted version thereof;
individually processing said first x-ray mammogram image and each of said at least one spatially inverted versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
12. The non-transitory computer-readable medium of claim 11 , wherein, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, said step of for comparing includes that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
13. The non-transitory computer-readable medium of claim 12 , wherein said at least one spatially inverted version includes (i) a vertically inverted version when said first x-ray mammogram is a CC view, and (ii) a horizontally inverted version when said first x-ray mammogram is an MLO view.
14. The non-transitory computer-readable medium of claim 11 , said foundational CAD processing algorithm also being characterized by non-rotational invariance, the computer program further performs the steps of: processing said first x-ray mammogram image to generate at least one rotated version thereof; individually processing each of said at least one rotated versions using said foundational CAD algorithm to generate at least one additional CAD detection set, wherein said step for comparing includes said at least one additional CAD detection set in said generation of said overall CAD detection set.
15. The non-transitory computer-readable medium of claim 14 , said foundational CAD processing algorithm also being characterized by non-shift invariance, the computer program further performs the steps of: processing said first x-ray mammogram image to generate at least one shifted version thereof; individually processing each of said at least one shifted versions using said foundational CAD algorithm to generate at least one further additional CAD detection set, wherein said step for comparing includes said at least one further additional CAD detection set in said generation of said overall CAD detection set.
16. The non-transitory computer-readable medium of claim 15 , wherein said step for comparing implements a voting technique in which said breast location is included in said overall CAD detection set only when at least a predetermined percentage of said CAD detection sets have identified said breast location for CAD marking.
17. The non-transitory computer-readable medium of claim 16 , the computer program further performs the step of displaying said overall CAD detection set in conjunction with said first x-ray mammogram image on an output display.
18. A method for computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a foundational CAD processing algorithm, said foundational CAD processing algorithm being characterized by non-shift-invariance, non-rotational-invariance, and non-inversional-invariance, the method comprising:
receiving a first x-ray mammogram image of a breast in digital form, wherein said first x-ray mammogram image comprises a CC view of the breast and said at least one spatially inverted version is inverted vertically in comparison to said first x-ray mammogram image;
processing by digital computer said first x-ray mammogram image to generate a plurality of altered versions thereof including at least one spatially inverted version thereof, at least one rotated version thereof, and at least one shifted version thereof; individually processing said first x-ray mammogram image and each of said plurality of altered versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
19. The method of claim 18 , wherein said comparing comprises, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, including that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
20. The method of claim 19 , wherein said collective evaluation of outcomes for said breast location comprises a voting technique in which said breast location is included in said overall CAD detection set only if at least seventy percent (70%) of said CAD detection sets have identified said breast location for CAD marking, and wherein the method further comprises displaying said overall CAD detection set to a viewer in conjunction with said first x-ray mammogram image on an output display.
21. The method of claim 18 , wherein said at least one rotated version includes at least ten (10) successively rotated versions of said first x-ray mammogram image having rotation angles distributed between about −5 degrees and 5 degrees.
22. The method of claim 18 , said first x-ray mammogram image having a number of pixels in a range of about 1K×1K to about 10K×10K pixels, wherein said at least one shifted version includes at least ten (10) successively shifted versions of said first x-ray mammogram image having shift amounts distributed between about 2 pixels and 20 pixels.
23. A method for computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a foundational CAD processing algorithm, said foundational CAD processing algorithm being characterized by at least one of non-shift-invariance, non-rotational-invariance, and non-inversional-invariance, the method comprising:
receiving a first x-ray mammogram image of a breast in digital form;
processing by digital computer said first x-ray mammogram image to generate at least one altered version thereof that differs therefrom by at least one of image shift, image rotation, and image inversion, wherein at least one of said altered versions comprises a spatially inverted version of said first x-ray mammogram image, and said first x-ray mammogram comprises an MLO view of the breast and said at least one altered version is inverted horizontally in comparison to said first x-ray mammogram image;
individually processing said first x-ray mammogram image and each of said at least one altered versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
24. The method of claim 23 , wherein said comparing comprises, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, including that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
25. The method of claim 24 , wherein said collective evaluation of outcomes for said breast location comprises a voting technique in which said breast location is included in said overall CAD detection set only if at least seventy percent (70%) of said CAD detection sets have identified said breast location for CAD marking, and wherein the method further comprises displaying said overall CAD detection set to a viewer in conjunction with said first x-ray mammogram image on an output display.
26. The method of claim 23 , wherein said at least one altered version includes at least ten (10) successively rotated versions of said first x-ray mammogram image having rotation angles distributed between about −5 degrees and 5 degrees.
27. The method of claim 23 , said first x-ray mammogram image having a number of pixels in a range of about 1K×1K to about 10K×10K pixels, wherein said at least one altered version includes at least ten (10) successively shifted versions of said first x-ray mammogram image having shift amounts distributed between about 2 pixels and 20 pixels.
28. A method for computer-aided detection (CAD) of anatomical abnormalities in x-ray mammograms based on a foundational CAD processing algorithm, said foundational CAD processing algorithm being characterized by non-shift-invariance, non-rotational-invariance, and non-inversional-invariance, the method comprising:
receiving a first x-ray mammogram image of a breast in digital form, wherein said first x-ray mammogram image comprises an MLO view of the breast and said at least one altered version is inverted horizontally in comparison to said first x-ray mammogram image;
processing by digital computer said first x-ray mammogram image to generate a plurality of altered versions thereof including at least one spatially inverted version thereof, at least one rotated version thereof, and at least one shifted version thereof; individually processing said first x-ray mammogram image and each of said plurality of altered versions using said foundational CAD algorithm to generate a respective plurality of individual CAD detection sets; and
comparing said plurality of CAD detection sets to generate an overall CAD detection set for said first x-ray mammogram image.
29. The method of claim 28 , wherein said comparing comprises, for each location in the breast identified for CAD marking in at least one of said CAD detection sets, including that breast location in said overall CAD detection set based on a collective evaluation of outcomes for that breast location in all of said CAD detection sets.
30. The method of claim 29 , wherein said collective evaluation of outcomes for said breast location comprises a voting technique in which said breast location is included in said overall CAD detection set only if at least seventy percent (70%) of said CAD detection sets have identified said breast location for CAD marking, and wherein the method further comprises displaying said overall CAD detection set to a viewer in conjunction with said first x-ray mammogram image on an output display.
31. The method of claim 28 , wherein said at least one rotated version includes at least ten (10) successively rotated versions of said first x-ray mammogram image having rotation angles distributed between about −5 degrees and 5 degrees.
32. The method of claim 28 , said first x-ray mammogram image having a number of pixels in a range of about 1K×1K to about 10K×10K pixels, wherein said at least one shifted version includes at least ten (10) successively shifted versions of said first x-ray mammogram image having shift amounts distributed between about 2 pixels and 20 pixels.